Xiangyu Zhao

Xiangyu Zhao

PhD Candidate

Imperial College London

About me

Hi! I am a third-year PhD student at the Department of Electrical and Electronic Engineering, Imperial College London, supervised by Dr Aaron Zhao. I am also a memeber of the DeepWok Lab. My PhD research topic is to explore learning beyond structured data, including topological deep learning on higher-order graphs, and graphs in foundation models. My long-term research interest is to push the ability of machine learning systems in learning more complex data structures, while reducing their reliance on specialised human knowledge on specific tasks. In June 2022, I graduated as an MEng in Computer Science from Trinity College, University of Cambridge, with a result of distinction, and was awarded the Senior Scholarship.

Outside of academics, I am a keen clarinettist, a 1st Dan kendoka, and a goalkeeper. I am a former member and soloist of the Cambridge University Chinese Orchestra Society, and performed in the orchestra’s Michaelmas Recitals and Annual Concerts in 2018, 2019, 2020 and 2022. I am also a member of the Imperial College Kendo Club and a former member of the Cambridge University Kendo Society, where I won the 2nd Place in the team matches in the 2020 University Taikai, and the 1st Place in the kyu-mixed individual matches in the 2023 Oxbridge Kendo Varsity matches.

Download my CV.

Interests
  • Graph Representation Learning
  • Topological Deep Learning
  • Multimodal Learning
  • Self-Supervised Learning
  • Reinforcement Learning
  • Generative Models
Education
  • PhD in Electrical and Electronic Engineering, 2023 –

    Imperial College London

  • MEng in Computer Science, 2021 – 2022

    University of Cambridge

  • BA in Computer Science, 2018 – 2021

    University of Cambridge

Publications

(2025). EC-Gate: Expansion Contribution-Aware Gating for Structure-Guided Message Passing in GNNs. In Proceedings of the 4th Learning on Graphs Conference (LoG 2025).

PDF Cite Project Poster

(2024). Enhancing Node Representations for Real-World Complex Networks with Topological Augmentation. In 27th European Conference on Artificial Intelligence (ECAI 2024).

PDF Cite Code Project Slides

(2023). Will More Expressive Graph Neural Networks do Better on Generative Tasks?. In Proceedings of the 2nd Learning on Graphs Conference (LoG 2023).

PDF Cite Code Project Poster

(2023). Hybrid Graph: A Unified Graph Representation with Datasets and Benchmarks for Complex Graphs. arXiv preprint arXiv:2306.05108.

PDF Cite Code Dataset Project

(2023). Task-Agnostic Graph Neural Network Evaluation via Adversarial Collaboration. 11th International Conference on Learning Representations (ICLR 2023) Machine Learning for Drug Discovery (MLDD) Workshop.

PDF Cite Code Project Poster

(2022). Towards a Competitive 3-Player Mahjong AI using Deep Reinforcement Learning. In 2022 IEEE Conference on Games (CoG).

PDF Cite Code Project Slides

(2022). Building a 3-Player Mahjong AI using Deep Reinforcement Learning. arXiv preprint arXiv:2202.12847.

PDF Cite Code Project